A cluster of servers may include two or more servers. Each server in the cluster provides a redundant backup information to another server in the cluster. Accordingly, server redundancy is desirable such that if a server in the cluster fails, another server in the cluster will automatically take over the failed server.
However, a large number of servers in a cluster may result in a large number of data being exchanged thereby impacting memory and network traffic. As such, servers in the cluster that are minimally accessed may take bandwith and memory space as the data from these servers is unnecessarily replicated to other servers in the cluster for redundancy. Therefore, what is needed is a technique for calculating a degree of participation of a server in a cluster to determine whether the server is to remain in the cluster.